AI News - Thursday, 12 November 2026 - Nvidia Board Jensen Huang, China Chip Export Rules, TSMC Anthropic OpenAI
In a Nutshell
Nvidia's board moved toward removing Jensen Huang just as Washington turned on the chipmaker, a sharp reminder that AI leadership can be repriced in a week. Elsewhere the platform layer kept moving: Gemini 4 Carbon shipped, DeepMind took on software proofs, and TSMC put a billion dollars behind Anthropic and OpenAI.
5-minute AI news update - 12 November 2026
In this AI News
Nvidia's board reportedly moves to oust CEO Jensen Huang as shares fall
[Industry] The board of the world's most valuable chipmaker is weighing removing a founder who has run it for 33 years, and the stock fell on the report. Any leadership change at Nvidia reshapes supply agreements, pricing and roadmap commitments that every AI programme depends on. Treat single-vendor compute plans as a live risk over the next two quarters. Source: TechCrunch
Huang affair reaches Washington as Congress weighs tighter China chip export rules
[Geopolitics] Lawmakers called for Huang's resignation and moved to tighten export controls on AI chips to China, an area where he had lobbied against restrictions. Export policy is now entangled with a personal scandal, which shortens the odds of abrupt rule changes. Institutions buying accelerated compute should model a stricter export regime, not the status quo. Source: CNBC
TSMC to invest $1 billion in Anthropic and OpenAI
[Funding] The world's largest chip foundry is taking equity in the two labs that buy its capacity, the clearest example yet of circular financing inside the AI boom. It ties lab funding, silicon supply and foundry revenue into one loop. That loop is efficient while demand grows and fragile if the AI trade keeps wobbling. Source: Ars Technica
DeepMind's AlphaGenome 2 predicts how mutations behave inside specific cell types
[Research] The new DNA model moves from reading the genome to modelling how individual cells interpret it, which is where disease actually starts. Better cell-level prediction shortens the path from variant to target. For any institution training life-science students, this is the tooling standard to teach against. Source: The Verge
DeepMind introduces AlphaProof Code to verify complex software at depth
[Research] Formal verification has been the bottleneck between AI-written code and code you can put in front of an auditor. A model that proves properties of large codebases changes what 'reviewed' means in engineering teams. It also raises a governance question: who signs off when the proof itself is machine-generated. Source: Google DeepMind
Gemini 4 Carbon ships to developers, enterprises, scientists and consumers
[Models] Google is pushing a single flagship across four very different audiences at once, which compresses the usual staged rollout. Enterprises get capability sooner but with less time to test guardrails. Plan an evaluation window before it lands in production tooling. Source: Google DeepMind
Gemini Robotics 2.5 targets real-world dexterity for general-purpose robots
[Models] Dexterity, not reasoning, is what keeps robots out of labs and warehouses. A model tuned for physical manipulation is the precondition for campus robotics pilots that actually hold up. Expect vendor demos to outrun your safety review, so write the review first. Source: Google DeepMind
Google launches a Gemini Decisions API for fast, calibrated multimodal decisions
[Tools] The pitch is calibrated confidence on high-volume decisions, which is exactly what automation in admissions, support and operations needs to be auditable. Calibration claims are the part to test yourself before trusting a threshold. It also competes directly with the decision-layer tooling most teams build in house. Source: Google
ChatGPT adds verified Business Agents, kept separate from personal conversations
[Tools] Separating a business agent from the user's own chat history is a data-boundary decision, not a UI detail, and it is the right one for institutional deployment. Verification of the business entity reduces impersonation risk. It makes third-party agent channels plausible on a university site without exposing student conversations. Source: The Verge
Anthropic adds Sheets, Whiteboards and Data artifacts, edging into office suites
[Tools] Claude is drifting from a chat product into a productivity suite with structured artifacts. That is the layer where institutional workflows actually live, and where vendor lock-in gets decided. Compare it against your current document stack before signing another multi-year seat deal. Source: The Register
United Airlines bars third-party AI agents from booking flights for travellers
[Policy] A major carrier drew the line at autonomous agents acting on a customer's behalf, citing liability and fraud exposure. The same argument will reach campus systems that let agents register students, file forms or move money. Identity delegation, not model quality, is the constraint to solve first. Source: The Verge
Google's Tabletop experiment lets users assemble custom collaborative AI apps
[Tools] Tabletop is a waitlisted Labs experiment for building small collaborative spaces around personal planning, a shape that maps well onto team workflows. It signals Google testing consumer-grade app assembly rather than another chatbot surface. Watch the access model before planning any pilot. Source: The Verge
CoreWeave jumps 24 percent after Q3 as it leans on prepayments and shorter contracts
[Funding] The neocloud is financing AI growth with customer cash up front instead of long leases, which shifts risk onto buyers. Shorter contracts cut both parties' exposure but raise renewal and price risk. If you rent GPU capacity, read the term sheet for prepayment clauses now. Source: CNBC
Britain's Fractile raises $720 million to build purpose-designed AI chips
[Funding] A European challenger reaching top-tier funding in AI silicon widens the supplier field beyond Nvidia and its immediate rivals. More competition matters most for buyers who want a second source. Watch whether the round converts into shipping silicon or into another design partnership. Source: TechCrunch
Meta loses Santosh Janardhan, its longtime head of infrastructure
[Industry] The person who built Meta's data-centre estate for over a decade is stepping down while the company spends at historic levels on AI capacity. Infrastructure leadership churn usually precedes a change in build strategy or a pause. It is a leading indicator for the whole colocation and power market. Source: TechCrunch
IEEE Spectrum: agentic AI is stretching zero-trust security past its design limits
[Research] Zero trust assumes identifiable principals making bounded requests, and autonomous agents cooperating with each other break both assumptions. The article documents how verification and accountability degrade as agents delegate. This is the architecture question any smart-campus agent rollout has to answer before launch. Source: IEEE Spectrum
Scott Wiener, Pelosi's likely successor, takes his AI regulation agenda to Congress
[Policy] Wiener wrote California's AI transparency laws and is now positioned to move similar rules federally. State-level precedent becoming federal drafting language is the pattern to watch. Compliance planning built around California requirements is the safer bet for the next cycle. Source: MIT Technology Review
Higher ed IT leaders shift from reactive fixes to planned AI governance and vendor risk
[Education] The report describes IT moving from firefighting to deliberate planning on vendor risk, AI governance and classroom design. That is the same transition a CIO office has to make visible to a board. It is a useful external benchmark for how peer institutions sequence the work. Source: EdTech Magazine
Two institutions use vibe coding to rebuild internal processes without a vendor
[Education] Staff are now shipping working internal tools from natural-language prompts, which changes the build-versus-buy calculation for administrative workflows. The same capability creates shadow IT and untested code in production. Pair the enthusiasm with a review path or the savings arrive as incidents. Source: EdTech Magazine
The world of AI is evolving at full speed.
Become a Fellow at university-365.com
Become Superhuman... In a world of AI... Prompt Smart, Prompt UP!
Comments